Open in new window / Try shogun cloud
--- Log opened Fri Aug 26 00:00:25 2011
blackburnwhere is CIA01:25
CIA-87shogun: Sergey Lisitsyn master * r420fce1 / (2 files): Plif fixes -
CIA-87shogun: Sergey Lisitsyn master * rc31b795 / (2 files): Improved CDistance parallel implementation -
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]01:39
-!- f-x [~user@] has joined #shogun01:49
-!- f-x` [~user@] has joined #shogun04:23
-!- f-x [~user@] has quit [Ping timeout: 260 seconds]04:25
CIA-87shogun: Soeren Sonnenburg master * r0a8fcea / src/interfaces/modular/modshogun_ignores.i : fix swig java syntax error? -
CIA-87shogun: Soeren Sonnenburg master * r759fabb / (2 files in 2 dirs): add make check-examples-ruby_modular target -
CIA-87shogun: Soeren Sonnenburg master * r443fa3e / (src/configure src/interfaces/modular/modshogun_ignores.i): comment the == ignore line as java_modular chokes on it -
@sonney2kserialhex, ping ping09:02
CIA-87shogun: Soeren Sonnenburg master * r14bccf1 / examples/undocumented/csharp_modular/ : improve to take arguments with scripts to be tested -
CIA-87shogun: Soeren Sonnenburg master * rb17f07e / src/interfaces/modular/modshogun_ignores.i : temporarily disable all ignores of operators -
CIA-87shogun: Soeren Sonnenburg master * ref7f6a3 / src/configure :10:45
CIA-87shogun: make configure fail when java is enabled but java/javac/ant are not10:45
CIA-87shogun: detected -
CIA-87shogun: Soeren Sonnenburg master * r3997e8d / src/Makefile.template : add check examples targets -
-!- shogun-buildbot [] has quit [Quit: buildmaster reconfigured: bot disconnecting]11:02
-!- shogun-buildbot [] has joined #shogun11:03
@sonney2kf-x`, are you around?11:07
f-x`sonney2k: yes11:12
-!- f-x [~user@] has joined #shogun11:13
@sonney2kf-x`/f-x, I just wanted to ask whether you think your submitted code is in releasable state11:13
@sonney2kor what you still want to improve11:14
@sonney2kas you might know we currently intend to release on aug 3111:14
f-x`sonney2k: it depends on what feature is being used11:15
f-xi.e. StreamingSimpleFeatures etc will mostly be good11:15
@sonney2kf-x, ok and which are not so?11:15
f-x`but there are others which I'll be happier if I work on before releasing11:15
f-xsonney2k: eg. StreamingStringFeatures is something I want to experiment on further11:16
@sonney2kf-x, is it doable for you to finish the tests within the next few days?\11:16
f-x`VW is good, one minor addition I can think of will make it good enough11:16
@sonney2kf-x, you can always improve11:17
@sonney2kthis won't be the last release of shogun11:17
f-x`sonney2k: yes, i think i can do it if i work on the testing part now11:17
@sonney2khowever it would be great if we could release sth rather now/soon and then gradually continue development11:17
@sonney2kso better fix/test now and mark things unstable if you are not convinced that they work11:17
@sonney2kthe shogun-buildbot is doing a great job in helping us detect regressions...11:18
f-x`sonney2k: ok, i'll concentrate on getting the *Features things working in a more stable manner first11:18
f-xone features class per day should mean all 3 covered in the next 3 days, which is hopefully okay11:19
@sonney2kyes ok11:19
@sonney2kit is best if you write some small (python_modular) test script for them11:20
@sonney2kthen these will always be tested11:20
f-x`sonney2k: could you point me to some similar scripts for the features classes we already have?11:20
@sonney2kf-x, just look at the python_modular/feature*.py examples11:20
f-x`ok, will do11:21
-!- f-x [~user@] has left #shogun ["ERC Version 5.3 (IRC client for Emacs)"]11:22
-!- f-x` [~user@] has quit [Quit: ERC Version 5.3 (IRC client for Emacs)]11:22
-!- f-x [~user@] has joined #shogun11:23
@sonney2kf-x, thank you very much!11:24
f-xsonney2k: np, will look to get it done as soon as possible11:24
@sonney2klets make this the best shogun release ever :D11:25
f-xyay! :)11:25
CIA-87shogun: Soeren Sonnenburg master * rcfbef75 / doc/Makefile : fix make doc -
-!- blackburn [~blackburn@] has joined #shogun12:06
blackburnsonney2k: so did you manage to fix java compilation? it works for me btw12:08
@sonney2kblackburn, yes - there were some things just not installed12:28
@sonney2kon the buildslave I mean12:28
@sonney2kblackburn, but look at
@sonney2kstill lots of errors...12:28
@sonney2kthough some of them seem to be easiliy solved12:29
@sonney2k error in classifier_liblinear_modular.java12:29
@sonney2kUnable to create matrix from .../data/fm_train_real.dat: .../data/fm_train_real.dat (No such file or directory)12:29
@sonney2kor should I say .........../data/ *lol*12:29
@sonney2kblackburn, btw csharp works more or less too now but also has issues running the examples12:30
blackburnI'll fix java soon12:30
@sonney2kthough I need sploving's help12:30
blackburnhe disappeared :D12:31
@sonney2kI hope he will do the fixes12:31
@sonney2kblackburn, do you know if serialhex has time to fix all the ruby examples?12:34
blackburnno idea12:34
@sonney2kthe build now also works but many new tests fail12:34
@sonney2kthe ones that are just not yet ported for real12:34
@sonney2kin the worst case we have to do it all...12:34
@sonney2kor remove non-working examples when we run out of time12:35
CIA-87shogun: Soeren Sonnenburg master * r4a4d755 / src/interfaces/modular/modshogun_ignores.i : turn on ignores again as the culprit was broken java detection in configure -
blackburnsonney2k: uh.. found bug13:28
blackburnit seems that if some getter is called and result isn't 'saved' anywhere - it leads to memory error13:29
blackburnin java13:29
@sonney2kresult isn't saved - what does that mean?13:31
@sonney2kf-x, please fix the if thing13:33
@sonney2kand then I commit13:33
@sonney2kblackburn, you mean a shogun object13:33
@sonney2kor double[,] ?13:34
blackburnsonney2k: double matrix13:36
blackburni.e. get_histogram()13:36
@sonney2kblackburn, please be more verbose13:37
f-xsonney2k: ah, i see - no if () needed before any SG_UNREF?13:37
@sonney2kI still don't understand13:37
@sonney2kf-x, SG_UNREF checks that already13:37
blackburnsonney2k: yeah I overvodkaed13:37
blackburnI will check before13:37
@sonney2kblackburn, it is more like you chat while riding on you bear13:38
blackburnsonney2k: checked, I was wrong13:38
f-xsonney2k: ok, done13:39
blackburnhistogram bug I guess13:39
CIA-87shogun: Shashwat Lal Das master * r66c4ffd / (3 files in 2 dirs): Improvements in StreamingStringFeatures and reindentation of ParseBuffer. -
CIA-87shogun: Shashwat Lal Das master * r65a020c / examples/undocumented/libshogun/streaming_stringfeatures.cpp : StreamingStringFeatures libshogun example. -
CIA-87shogun: Shashwat Lal Das master * ra6bf023 / examples/undocumented/libshogun/streaming_stringfeatures.cpp : Commented StreamingStringFeatures example better. -
CIA-87shogun: Shashwat Lal Das master * r1d81f89 / src/shogun/features/StreamingStringFeatures.h : Removed pointer checking before SG_UNREF in StreamingStringFeatures. -
CIA-87shogun: Soeren Sonnenburg master * r58b557f / (4 files in 3 dirs):13:40
CIA-87shogun: Merge pull request #311 from frx/streaming_vw13:40
CIA-87shogun: StreamingStringFeatures improvements -
@sonney2kf-x, ^ merged13:40
f-xok :13:40
f-xsome more coming up soon13:40
blackburnfor (i=0; i< (int32_t) (1<<16); i++)13:42
blackburnokay not a bug13:43
blackburnsonney2k: histogram get_histogram fails in java..13:48
blackburnno idea why13:48
blackburnsonney2k: STACK OVERFLOW!13:52
@sonney2kblackburn, uhh13:52
@sonney2khow that and where13:52
blackburnget_histogram returns vector of size 6553613:52
blackburnso distribution_histogram_modular fails13:53
@sonney2kbut the vector is not on stack?!13:53
blackburnlooks like it is13:53
@sonney2kblackburn, you talk about SGVector ?13:53
blackburnyes, SGVector mapped to DoubleMatrix13:54
@sonney2kbut SGVector is just a double* and int13:54
@sonney2knot sure about the java typemap mapping13:54
blackburnI tried return just 40k of histogram elements13:54
blackburnand got stack overflow13:54
blackburnif it is 65536 all is fucked up with SIGSERV13:55
blackburnsonney2k: JNITYPE arr[cols]13:56
@sonney2kblackburn, indeed loock at swig_typemaps.i line 11313:56
@sonney2kso please do a SG_MALLOC there13:56
blackburn*** sploving13:57
@sonney2kblackburn, same thing will happen in SGMatrix13:58
blackburnsonney2k: get_degree defined two times in SpectrumMismatchRBFKernel :D14:07
blackburnsonney2k: okay fixed14:15
CIA-87shogun: Sergey Lisitsyn master * r02feda6 / src/shogun/distributions/Histogram.cpp : Fixes for Histogram -
CIA-87shogun: Sergey Lisitsyn master * r5630a1d / (2 files): Removed unnecessary double definitions of methods at string kernels -
CIA-87shogun: Sergey Lisitsyn master * r63b5aec / src/interfaces/java_modular/swig_typemaps.i : Changed memory allocation method in java typemaps -
CIA-87shogun: Sergey Lisitsyn master * r9c4972b / (3 files): Some java examples fixes -
@sonney2kblackburn, btw now we have to check that this is not causing a memory leak14:24
blackburnsonney2k: no, shouldn't14:24
blackburnsonney2k: java would crash with some MTHFCKR SIGNAL :D14:25
@sonney2kblackburn, I mean you might have to call SG_FREE later...14:26
@sonney2kblackburn, so do all java examples work now?14:28
@sonney2kblackburn, wait you have a copy'n'paste bug here14:29
@sonney2kline 48014:29
blackburnsonney2k: no, not all14:31
blackburnin progress14:31
blackburnsonney2k: but where to call SG_FREE??14:31
@sonney2kafter the JNI_CALL thing?14:32
blackburnwill it copy array?14:32
@sonney2kI don't know14:32
blackburnme too14:32
blackburnsonney2k: looks like no14:34
blackburnit uses some Set[Type]ArrayRegion14:34
@sonney2kSet<PrimitiveType>ArrayRegion Routines14:35
@sonney2kit only copies14:35
@sonney2kso we have to SG_FREE14:35
CIA-87shogun: Sergey Lisitsyn master * r911732f / src/interfaces/java_modular/swig_typemaps.i : Fixes for java typemaps -
blackburnsorry gtg14:43
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]14:43
-!- sploving1 [~sploving@] has joined #shogun14:53
sploving1sonney2k, are you around?14:53
@sonney2ksploving1, yes14:53
sploving1hist[0]=92 what do you mean14:54
sploving1I did not understand it clearly14:54
@sonney2ka histogram is computed over the characters in the string14:54
@sonney2kthat says that the string null termination character appeared 92 times14:54
@sonney2k(I think you have 92 strings there)14:54
sploving1what does get_num_bits_in_histogram()=3  mean?14:55
@sonney2kthat one would need 3 bits to encode the 5 different characters14:55
@sonney2kbut DNA has only 4 characters - 2bits needed so that won't fit14:56
sploving1 even though a DNA alphabet was expected so only ord('A') (C/G/T) should14:57
@sonney2kbut you have a '\0' in the string14:58
sploving1you mean ACGT'\0'14:58
@sonney2kI guess that you pass a string 1 byte too long14:58
sploving1sonney2k, in C we have '\0' in the string14:58
@sonney2ksploving1, in shogun not14:58
@sonney2kwe have a ptr and a length14:59
sploving1the length is not include the '\0', so I add 1 in the typemap14:59
@sonney2kso one fix would be to just say str.length = len-114:59
sploving1str.length mean shogun length, len means C length, am I right?15:00
@sonney2kthat is the problem then15:00
@sonney2kyes len means csharp len15:00
sploving1max_len = shogun::CMath::max(len, max_len); strings[i].slen = len + 1;15:01
sploving1maybe this is where is the but15:01
sploving1max_len = shogun::CMath::max(len, max_len); strings[i].slen = len;15:01
sploving1am I right?15:01
@sonney2kcould be15:02
@sonney2klikely yes15:03
sploving1I will have a try15:03
@sonney2ksploving1, OK - when that is done could you please have a look at the gaussian kernel example? it doesn't return the correct results (looks more like uninited memory...)15:04
sploving1okay. I saw your email. but have no idea about that now15:04
sploving1why the other examples work well15:04
sploving1but gaussian_kernel not work well15:05
@sonney2ksploving1, the linear kernel one also crashes15:06
@sonney2kseems to be the same problem15:06
sploving1fixed the stringlist but and pushed15:09
CIA-87shogun: Baozeng Ding master * r1c36dfa / src/interfaces/csharp_modular/swig_typemaps.i : fix stringlist bug -
CIA-87shogun: Soeren Sonnenburg master * r2226301 / src/interfaces/csharp_modular/swig_typemaps.i :15:11
CIA-87shogun: Merge pull request #312 from sploving/master15:11
CIA-87shogun: fix stringlist bug -
@sonney2ksploving1, thanks15:11
sploving1I need a simple example to test the matrix15:11
@sonney2ksploving1, the gaussian kernel one15:11
sploving1when I output the gaussian_kernel, it prints a lot of numbers15:11
sploving1I did not even know which is in the triangle15:12
sploving1linear is easier?15:12
sploving1not at all15:12
sploving1foreach(double item in km_train) {15:13
sploving1maybe I just printf the first line15:13
sploving1the first element should be 1, am I right?15:13
@sonney2ksploving1, linear crashes here15:13
@sonney2kwhy not use kernel_gaussian_modular.cs?15:15
@sonney2kyou can simplify gaussian if you want15:15
@sonney2kjust use15:15
sploving1I think kernel_gaussian_modular.cs is easy15:15
sploving1why you say k[i,i]=115:15
@sonney2k        GaussianKernel kernel = new GaussianKernel(feats_train, feats_train, width);15:15
@sonney2k        double[,] km_train = kernel.get_kernel_matrix();15:15
sploving1this is shogun inner issue?15:15
@sonney2kthen km_train[i,i] should be 1.015:15
@sonney2ksploving1, by definition of the gaussian kernel15:16
@sonney2kexp(0) == 115:16
sploving1you mean Console.Write(km_train[0,0])  should be 115:16
@sonney2ki,i I would use though15:16
@sonney2kbut yes15:17
@sonney2kkm_train[0,0] should be 1 too :)15:17
sploving1km_train[0,0]  is 1 here15:19
sploving1I will test km_train[1,1]15:19
sploving1sonney2k, in my machine it is 115:21
* sonney2k checks again15:21
sploving1I test a random number km_train[8,8]15:22
sploving1it print 115:22
@sonney2ksploving1, uhh it could be a printing problem15:22
@sonney2kbecause there are no spaces between numbers15:23
sploving1I am not very familiar with Csharp15:23
sploving1another issue, is linear_word??15:24
@sonney2ksploving1, yeah but if you replace GaussianKernel in the example with LinearKernel - it already fails15:24
@sonney2k(remove width parameter)15:24
sploving1LinearKernel is different from gaussian kernel15:27
sploving1LinearKernel kernel = new LinearKernel(feats_train, feats_test);15:27
sploving1not LinearKernel kernel = new LinearKernel(feats_train, feats_train)15:28
@sonney2kerr feats_train, feats_train15:28
@sonney2kthen feats_train, feats_test15:28
@sonney2kwill be two different matrices at the end15:28
@sonney2khmmhh just changing this in the gaussian_modular example didn't cause any trouble15:30
sploving1then what do you mean fails?15:31
sploving1what is the output?15:33
sploving1it works in my machine15:33
sploving1I just run it15:34
@sonney2k error in kernel_linear_word_modular.cs15:34
@sonney2k  at (wrapper managed-to-native) System.Runtime.InteropServices.Marshal.copy_from_unmanaged (intptr,int,System.Array,int) <0x0005c>15:34
@sonney2k  at (wrapper managed-to-native) System.Runtime.InteropServices.Marshal.copy_from_unmanaged (intptr,int,System.Array,int) <0x0005c>15:34
@sonney2k  at System.Runtime.InteropServices.Marshal.Copy (intptr,double[],int,int) <0x0002b>15:34
@sonney2k  at Kernel.get_kernel_matrix () <0x00117>15:34
@sonney2k  at kernel_linear_word_modular.Main () <0x00417>15:34
@sonney2k  at (wrapper runtime-invoke) object.runtime_invoke_void (object,intptr,intptr,intptr) <0x00049>15:34
@sonney2kNative stacktrace:15:34
@sonney2kmono() [0x487ad6]15:34
@sonney2kmono() [0x4d354f]15:34
@sonney2k/lib/ [0x2ae01a083f60]15:34
@sonney2k/lib/ [0x2ae01a5938ab]15:34
@sonney2kDebug info from gdb:15:34
@sonney2k82../sysdeps/unix/syscall-template.S: No such file or directory.15:34
@sonney2k[Thread debugging using libthread_db enabled]15:34
@sonney2k[New Thread 0x2aaaabfc0700 (LWP 27816)]15:34
@sonney2k[New Thread 0x2aaaaacac700 (LWP 27815)]15:34
@sonney2k0x00002ae01a0830bd in read () at ../sysdeps/unix/syscall-template.S:8215:34
@sonney2kin ../sysdeps/unix/syscall-template.S15:34
@sonney2k  3 Thread 0x2aaaaacac700 (LWP 27815)  pthread_cond_wait@@GLIBC_2.3.2 ()15:34
@sonney2k    at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:16215:34
@sonney2k  2 Thread 0x2aaaabfc0700 (LWP 27816)  sem_wait ()15:34
@sonney2k    at ../nptl/sysdeps/unix/sysv/linux/x86_64/sem_wait.S:8615:34
@sonney2k* 1 Thread 0x2ae01aaa6560 (LWP 27814)  0x00002ae01a0830bd in read ()15:34
@sonney2k    at ../sysdeps/unix/syscall-template.S:8215:35
@sonney2kThread 3 (Thread 0x2aaaaacac700 (LWP 27815)):15:35
@sonney2k#0  pthread_cond_wait@@GLIBC_2.3.2 ()15:35
sploving1sploving@sploving-laptop:~/shogun/examples/undocumented/csharp_modular$ gmcs kernel_linear_word_modular.cs ../../../src/interfaces/csharp_modular/*.cs Load.cs15:35
@sonney2k    at ../nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S:16215:35
@sonney2k#1  0x00000000005b9923 in ?? ()15:35
@sonney2k#2  0x00000000005c3df5 in ?? ()15:35
@sonney2k#3  0x00000000005ba6cc in ?? ()15:35
@sonney2k#4  0x00002ae01a07b8ba in start_thread (arg=<value optimized out>)15:35
@sonney2k    at pthread_create.c:30015:35
@sonney2k#5  0x00002ae01a5e302d in clone ()15:35
@sonney2k    at ../sysdeps/unix/sysv/linux/x86_64/clone.S:11215:35
@sonney2k#6  0x0000000000000000 in ?? ()15:35
@sonney2kThread 2 (Thread 0x2aaaabfc0700 (LWP 27816)):15:35
sploving1gmcs kernel_linear_word_modular.cs ../../../src/interfaces/csharp_modular/*.cs Load.cs15:35
sploving1here it print out the above numbers15:36
sploving1maybe others can help have a try15:37
@sonney2ksploving1, how many cores do you have?15:37
sploving12 core15:37
@sonney2kme too15:38
@sonney2kcore2 duo15:38
sploving1I think others can help have a try15:38
@sonney2ksploving1, it is also crashing on the build bot15:40
sploving1how to run build bot?15:42
sploving1buid bot is ??15:42
@sonney2ksploving1, as soon as you commit sth, it compiles everything and checks the examples15:44
@sonney2ksploving1, 64bit machine or 32bit?15:47
sploving1sonney2k, can I run it in my machine?15:47
@sonney2kyours I mean?15:47
sploving1you mean OS?15:47
@sonney2kI think the out typemap is broken15:47
@sonney2kbuildbot and my notebook here are 64bit15:48
@sonney2kso I guess that is the reason then15:48
* sonney2k checks the out typemap for SGMatrix15:48
sploving1soney2k, buildbot is another machine? super machine?15:48
sploving1I am so interested in it15:48
@sonney2ksploving1, look at line 13015:49
@sonney2kin swig_typemaps.i15:49
sploving1res[i + 2] = (CTYPE) $1.matrix[i];15:49
@sonney2kI mean what is CTYPE?15:49
@sonney2ksth like int?15:49
@sonney2kyou then first assign res[0]  / res[1] with some lengths (integers)15:50
@sonney2kand then res[i+2] with the ptr to data I guess?15:50
@sonney2kon my machine ptrs will be 64 bit I think so this could be the reason15:50
sploving1int, long, float, double, short are all CTYPE15:50
@sonney2ksploving1, buildbot is some core i7 2600 w/ 16G memory just building shogun all the time15:51
sploving1just a core? not a machine / notebook with disk?15:52
@sonney2kintel core i715:52
@sonney2kthe processor15:52
@sonney2kjust some machine15:52
sploving1oh. I know, you just named a popular machine, which is running the shogun always15:53
sploving1are there any way we can run shogun in irc?15:54
@sonney2kshogun-buildbot, dance15:54
sploving1I saw lua irc can run lua command15:54
@sonney2ksploving1, back to the SGMatrix out typemap15:55
@sonney2kcan't you use the struct SGMatrix has?15:55
@sonney2kand then in the csout typemap just use the respective ptrs15:55
sploving1not understand15:56
@sonney2ksploving1, what you did doesn't work15:56
@sonney2kyou allocate an array of e.g. uint16_t15:57
@sonney2kand then assign rows / cols in elements 0/115:57
@sonney2kand the data ptr to index 215:57
@sonney2know when e.g. rows is bigger than 65536 rows etc won't fit in that datatype15:58
@sonney2knor the 32bit (in your case) or 64bit data ptr15:58
@sonney2kso why not use the SGMatrix struct directly?15:58
sploving1%typemap(imtype, out="IntPtr", inattributes="int rows, int cols, [MarshalAs(UnmanagedType.LPArray)]") shogun::SGMatrix<SGTYPE>%{CSHARPTYPE[,]%}15:59
sploving1in Csharp, we cannot know rows and cols16:00
sploving1so I mange the typemap to add two args, rows and cols16:01
@sonney2kis the LPArray a ptr to the array?16:01
@sonney2kadding these two args is fine - but you need a big enough buffer to hold them16:01
@sonney2kso CTYPE cannot work16:01
-!- CIA-87 [] has quit [Ping timeout: 252 seconds]16:01
-!- CIA-3 [] has joined #shogun16:02
@sonney2kyou need sth that is 2*size(int32_t)+size(void*) big and then assign sizes to the respective elements16:03
@sonney2kactually it is not a ptr16:03
@sonney2kso sizeof(CTYPE)*num_rows*num_cols + 2*sizeof(int32_t)16:04
@sonney2ksploving1, do you agree?16:04
sploving1let me think16:05
f-xsonney2k: do you think it is ok if i typedef something like vw_size_t to uint32_t and use it in place of size_t in vw?16:12
@sonney2ksploving1, any idea how to properly fix it?16:12
shogun-buildbotbuild #118 of r_static is complete: Failure [failed test test_1]  Build details are at  blamelist: sploving1@gmail.com16:12
sploving1sonney2k, how to use SG_MALLOC to malloc a two int and len*sizeof(CTYPE)16:12
@sonney2kf-x, sure16:13
f-xsonney2k: ok, i'll see how that works out16:13
@sonney2kf-x, but why not uint32_t directly (or whaever was meant?)16:13
@sonney2ksploving1, only possilbe with hacks16:13
f-xsonney2k: i feel something like *size_t would explain the meaning better16:13
f-xif we convert it to uint32_t we lose out on some info about the intended use of the variable16:14
@sonney2kf-x, ok16:14
@sonney2ksploving1, that is why I suggested a struc16:14
@sonney2ksploving1, you could use16:15
@sonney2kchar* res = SG_MALLOC(char, sizeof(CTYPE)*len + 2*sizeof(int32_t));16:15
@sonney2k    ((int32_t*) res)[0] = rows;16:15
@sonney2k    ((int32_t*) res)[1] = cols;16:15
@sonney2k    CTYPE* dst = (CTYPE*) (res+2*sizeof(int32_t));16:15
@sonney2k    for (i = 0; i < len; i++) {16:15
@sonney2k        dst[i + 2] = (CTYPE) $1.matrix[i];16:15
@sonney2k    }16:15
@sonney2kor so16:15
@sonney2kit seems $result is a CTYPE* though16:16
@sonney2kso you might have to cast that again16:16
@sonney2kand then when you marshal do the same (un)mangling16:16
sploving1sonney, souds right. but dst is no need. res[i+2] is OK16:18
@sonney2kno res[i+2] is not16:18
@sonney2kif we have16:18
@sonney2kchar* res;16:18
@sonney2kthen res[i+2] would be the 3rd byte16:19
@sonney2kand only one byte for rows/cols16:19
sploving1CTYPE* dst = (CTYPE*) (res+2*sizeof(int32_t)); why?16:20
sploving1CTYPE* dst = (CTYPE*) res16:20
sploving1is oK16:20
sploving1then we use des[i+2]16:20
@sonney2ksploving1, rows and cols are always integers - correct?16:22
@sonney2kso they are 4 bytes always no matter what CTYPE we use - right?16:22
sploving1you mean another way?16:23
@sonney2kso when CTYPE==int or then dst[i+2] would point at byte 9 - which is correct16:23
@sonney2kbut when CTYPE==char then dst[i+2] points to byte 3 - > kaboom16:24
@sonney2kdo you see the problem?16:24
sploving1i know, then it should be  dst[i] = (CTYPE) $1.matrix[i];16:25
sploving1not dst[i + 2] = (CTYPE) $1.matrix[i];16:25
sploving1I think16:25
@sonney2kdon't you want to marshal the following sequence:16:26
@sonney2kint,int, CTYPE[some_size]16:26
@sonney2kahh dst[i] yes not res[i+2]16:26
-!- f-x` [~user@] has joined #shogun16:27
shogun-buildbotbuild #113 of python_static is complete: Failure [failed test_1]  Build details are at  blamelist: sploving1@gmail.com16:28
@sonney2kwhat's wrong with the buildbot now?!16:30
-!- f-x` [~user@] has quit [Client Quit]16:30
-!- f-x [~user@] has quit [Ping timeout: 260 seconds]16:30
-!- f-x [~user@] has joined #shogun16:30
@sonney2ksploving1, can you fix it?16:40
shogun-buildbotbuild #125 of cmdline_static is complete: Failure [failed test_1]  Build details are at  blamelist: sploving1@gmail.com16:42
CIA-3shogun: Shashwat Lal Das master * r54bc14c / (21 files in 4 dirs): Conversion of vw's size_t to typedef'd uint32_t -
CIA-3shogun: Soeren Sonnenburg master * r653e96c / (21 files in 4 dirs):16:45
CIA-3shogun: Merge pull request #313 from frx/streaming_vw16:45
CIA-3shogun: Conversion of vw's size_t to typedef'd uint32_t -
shogun-buildbotbuild #117 of octave_static is complete: Failure [failed test_1]  Build details are at  blamelist: sploving1@gmail.com16:45
sploving1what is the matter?16:47
@sonney2ksploving1, it is crashing because of the wrong assignments / array sizes16:48
@sonney2ksploving1, just create a ByteFeatures example and you will see also on your machine16:49
sploving1maybe there is something missing16:50
sploving1when I use len instead of len+116:50
sploving1sonney2k,  cannot convert 'char*' to 'double*' in assignment16:52
sploving1can I use void * res = SG_MALLOC(char, sizeof(CTYPE) * len + 2 * sizeof(int32_t));16:52
sploving1instead of char *16:52
@sonney2ksploving1, you can also cast at the end16:52
sploving1when I use CTYPE *r = (CTYPE *)res;  it means doulbe * r = (double*) res16:53
sploving1cannot work16:53
sploving1as res is char*16:53
@sonney2kserialhex, but you can use $result = (CTYPE*) res;16:54
@sonney2ksploving, but you can use $result = (CTYPE*) res;16:54
@sonney2khowever you need to make sure that the marshaling is ok16:54
@sonney2kbecause it is in fact not a (CTYPE*) but some *whatever*16:54
sploving1$rest is not double*16:55
sploving1$result is not doulbe *16:55
sploving1$result can use any cast, you mean16:55
@sonney2kwhat is result?16:55
@sonney2k(CTYPE*) yes16:55
@sonney2ksploving1, but you also need to adjust the csout typemap16:56
sploving1I know that16:56
@sonney2ksploving1, ok then please do - thanks16:59
shogun-buildbotbuild #120 of octave_modular is complete: Failure [failed compile]  Build details are at  blamelist: sonne@debian.org17:01
sploving1sonney2k, pulled17:07
shogun-buildbotbuild #131 of libshogun is complete: Failure [failed compile]  Build details are at  blamelist: fraggerx@gmail.com17:19
@sonney2ksploving1, btw is that bug also in vectors?17:20
CIA-3shogun: Baozeng Ding master * reb8b842 / src/interfaces/csharp_modular/swig_typemaps.i : fix rows and cols type bug -
CIA-3shogun: Soeren Sonnenburg master * rfcf7316 / src/interfaces/csharp_modular/swig_typemaps.i :17:20
CIA-3shogun: Merge pull request #314 from sploving/master17:20
CIA-3shogun: fix rows and cols bug -
sploving1sonney2k, fixed toghter17:21
@sonney2kgreat - thanks!17:22
@sonney2ksploving1, for strings you use sth else right?17:22
sploving1sonney2k, about the stringlist, I took carefully have a check, but did not find any where to fix17:23
@sonney2kf-x, you forgot to add shogun/classifier/vw/vw_constants.h:17:23
@sonney2kf-x, it fails to build now17:23
sploving1I gtg bye.17:24
-!- sploving1 [~sploving@] has left #shogun []17:24
shogun-buildbotbuild #121 of octave_modular is complete: Success [build successful]  Build details are at
f-xsonney2k: oh, in the .i? sorry17:35
@sonney2kf-x, not the actual .h17:35
f-xit compiled on mine, checking again17:35
f-xi understand - git add17:36
@sonney2kf-x, yes17:36
@sonney2knot sure if you forgot it in the .i file too17:36
@sonney2kthen it will also break there17:36
f-xsonney2k: will add it there as well17:36
@sonney2kthe faster the better :)17:37
f-xsonney2k: which .i file would be a good place to add it? i think it should work without the .i inclusion though17:38
f-xit doesn't provide any functions for the target language as such17:38
@sonney2kthen not needed17:38
@sonney2kpull request please17:38
f-xjust a sec17:39
f-xthere you go17:40
CIA-3shogun: Shashwat Lal Das master * r3dbfc26 / src/shogun/classifier/vw/vw_constants.h : Added vw_constants.h to repository. -
CIA-3shogun: Soeren Sonnenburg master * r3afc64d / src/shogun/classifier/vw/vw_constants.h :17:41
CIA-3shogun: Merge pull request #315 from frx/streaming_vw17:41
CIA-3shogun: Added vw_constants.h to repository. -
shogun-buildbotbuild #133 of libshogun is complete: Success [build successful]  Build details are at
@sonney2kf-x, see ^17:42
@sonney2kall good again17:42
f-xphew :)17:42
CIA-3shogun: Soeren Sonnenburg master * r148b751 / src/interfaces/csharp_modular/swig_typemaps.i : fix int_32_t typo -
shogun-buildbotbuild #135 of libshogun is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #127 of cmdline_static is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #120 of r_static is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #119 of octave_static is complete: Success [build successful]  Build details are at
shogun-buildbotbuild #115 of python_static is complete: Success [build successful]  Build details are at
-!- blackburn [~blackburn@] has joined #shogun18:27
blackburnsonney2k: uh what a nice error to debug18:28
blackburncan you please comment out plotting thing in and run it?18:30
CIA-3shogun: Sergey Lisitsyn master * raa95573 / src/shogun/mathematics/lapack.cpp : Remove asserts from DSYGVX -
CIA-3shogun: Sergey Lisitsyn master * ra920d11 / src/shogun/preprocessor/LaplacianEigenmaps.cpp : More informative error checking in laplacian eigenmaps -
-!- f-x` [~user@] has joined #shogun19:26
-!- f-x [~user@] has quit [Ping timeout: 260 seconds]19:27
@sonney2kblackburn, do the other graphical examples work for you?19:29
@sonney2kdoes anyone here have a 32bit system19:46
@sonney2kf-x`, blackburn you?19:46
blackburnsonney2k: both and works for me19:50
blackburnsonney2k: yes I have x86 system19:50
blackburnsonney2k: could you please comment out lines 53,54 in and run?19:52
blackburnah and 5019:52
@sonney2kblackburn, could you please compile the csharp_modular interface with swig_typemaps.i line 154 changed like20:52
@sonney2k-       Marshal.Copy(new IntPtr(ptr.ToInt32() + 2 * Marshal.SizeOf(typeof(int))), ret, 0, len);20:52
@sonney2k+       Marshal.Copy(new IntPtr(ptr.ToInt64() + 2 * Marshal.SizeOf(typeof(int))), ret, 0, len);20:52
blackburnwill try20:52
@sonney2kand then run make check-examples20:53
blackburnsonney2k: when you try my example ;)20:53
@sonney2kblackburn, compiling already20:54
@sonney2ktoo bad that my mono here is not yet .NET 4.0 compatible20:55
@sonney2kthis pointer hacking is beyond all my imagination :)20:55
@sonney2kblackburn we are damn close to c# stuff working reliably20:57
blackburn??? ??? ??????20:57
blackburnsonney2k: I sent some email to andreas, any chance he will answer during this weekend?20:57
blackburn(10:56:33 PM) blackburn: sonney2k: I sent some email to andreas, any chance he will answer during this weekend?20:57
blackburngot some connection problem20:58
blackburnnot sure it was delivered20:58
@sonney2kyes 2 times20:58
blackburnok sorry20:58
@sonney2klittle chance20:58
blackburnbad bad20:58
@sonney2kbtw, he is *the* ICA guy20:58
blackburnindependent component analysis?20:58
@sonney2kand he was very excited about what we did - maybe next year for gsoc he could mentor ica etc algorithms20:59
blackburnthere is a chance I implement it during our mid-season period20:59
blackburnokay I simply will send him a letter if all ok21:00
blackburnI asked him for some more information21:00
@sonney2kblackburn, I would suggest to ask him for relevant algorithms21:00
@sonney2khe is *the expert*21:00
@sonney2kblackburn, when I comment the lines above some figure pops up21:00
blackburnnice, could you please make a screenshot?21:01
blackburnbtw do you have arpack?21:01
@sonney2knumbers on axis overlap badly21:01
@sonney2kblackburn, sure21:01
blackburnsure about arpack?21:01
blackburnI have no solution for overlapping yet21:01
@sonney2kgrep -i arpack .config21:02
@sonney2kPOSTLINKFLAGS=  -lm -pthread -lhdf5 -ljson   -lxml2   -llapack -lcblas -llapack_atlas -larpack -lglpk -lm -llzo2 -lz -lbz2 -llzma21:02
blackburnplease disable arpack if it is enabled21:02
blackburnbug is in the lapack21:02
@sonney2klapack arpack?21:02
blackburnandreas used lapack when he got a bug21:02
blackburnnot arpack21:02
@sonney2kblackburn, ok disabling arpack21:03
@sonney2kblackburn, does c# do sth?21:03
blackburnI was installing mono develop files21:03
@sonney2kblackburn, is not working - can that be?21:05
@sonney2kblackburn, did you forget to commit ../../../../data/toy/hemisphere_color.dat21:05
blackburnsonney2k: don't you have hemisphere_color?21:08
blackburnI have..?21:08
@sonney2kblackburn, did you forget to git add it?21:09
blackburnit is here!21:09
@sonney2kblackburn, what is current?21:10
@sonney2k7ebf99a3944e92e157d934c8f11b6e5e5a27844d ?21:10
blackburndata is 90a72b3ec2df44dceb7d7d24c8bf878a257a786721:11
blackburnyou are one commit before21:11
@sonney2kI thought git submodule update will sync it21:13
@sonney2kobviously not21:13
@sonney2kanyway example doesn't run because I don't have 3d projection21:14
blackburndisable it21:14
blackburnerr comment21:14
@sonney2kblackburn, does all_evaluation still work  / is this still supported?21:14
blackburnall evaluation?21:15
blackburnno it fails on classify21:15
@sonney2kLocalTangentSpaceAlignment done21:15
@sonney2kTraceback (most recent call last):21:15
@sonney2k  File "", line 63, in <module>21:15
@sonney2k    new_feats = preproc.apply_to_feature_matrix(features)21:15
@sonney2k  File "/usr/local/lib/python2.6/dist-packages/", line 41669, in apply_to_feature_matrix21:15
@sonney2k    return _modshogun.SimpleRealPreprocessor_apply_to_feature_matrix(self, *args)21:15
@sonney2kSystemError: [ERROR] assertion *info==0 failed in file mathematics/lapack.cpp line 37321:15
@sonney2kblackburn, even when fixing this it doesn't work21:16
blackburnbut good21:16
blackburnsonney2k: it is the same error andreas got21:16
@sonney2kblackburn, I just wanted to know if you roc/prc things are making this obsolete or not21:16
@sonney2kgreat huh?21:16
blackburnno, all_evaluation won't work this way21:16
blackburnmany modifications to do21:16
blackburnsonney2k: your git is not up-to-date!21:17
@sonney2kblackburn, so did the c# compile?21:17
blackburnit did21:17
blackburnrunning tests21:17
@sonney2kTraceback (most recent call last):21:18
@sonney2k  File "", line 63, in <module>21:18
@sonney2k    new_feats = preproc.apply_to_feature_matrix(features)21:18
@sonney2k  File "/usr/local/lib/python2.6/dist-packages/", line 41669, in apply_to_feature_matrix21:18
@sonney2k    return _modshogun.SimpleRealPreprocessor_apply_to_feature_matrix(self, *args)21:18
blackburnall ok21:18
@sonney2kSystemError: [ERROR] DSYGVX failed with code: -20.21:18
@sonney2kblackburn, you sure that you have a 32bit system?21:18
blackburnsonney2k: yeah I'm pretty sure21:18
@sonney2kuname -m21:18
@sonney2kthen I have the ultimate(tm) fix :)21:19
@sonney2kblackburn, btw buildbot uptime  21:19:19 up 1 day,  1:32,  1 user,  load average: 0.13, 0.25, 0.5521:19
blackburnsonney2k: all examples are ok21:19
blackburnsonney2k: I guess my issue is x64 related too21:20
@sonney2kblackburn, that was with the Int64 stuff already in the swig_typemap?21:20
@sonney2k Marshal.Copy(new IntPtr(ptr.ToInt64() + 2 * Marshal.SizeOf(typeof(int))), ret, 0, len);21:21
blackburnsonney2k: yes I did suggested fix21:21
@sonney2k(Int32 on x86_64 will crash)21:21
@sonney2kblackburn, dammed! why did cwidmer let you pass - I explicitly told him to not do so ;-)21:27
CIA-3shogun: Soeren Sonnenburg master * rd8986e9 / src/interfaces/csharp_modular/swig_typemaps.i : use int64 to represent c#/c ptr -
CIA-3shogun: Soeren Sonnenburg master * rc5e08d9 / examples/undocumented/libshogun/mathematics_arpack.cpp : ifdef HAVE_ARPACK example to not fail on buildbot test -
blackburnsonney2k: yeah probably I shouldn't hehe21:33
CIA-3shogun: Soeren Sonnenburg master * r301e5f5 / (2 files): also ifdef HAVE_LAPACK liblinear based examples -
blackburnsonney2k: probably I need you to test some things21:38
@sonney2kblackburn, I really hope you are getting the remaining java examples fixed21:38
blackburnsonney2k: yeah sure sure but I need to fix laplacian shit21:39
@sonney2kit is only 4 java examples that fail now ...21:39
@sonney2kso shoot blackburn21:39
@sonney2kI am your man21:39
@sonney2kbtw did you hear anything from serialhex?21:39
blackburnsonney2k: anything like?21:40
@sonney2kI mean regarding ruby examples?21:40
blackburnsonney2k: mathematics/lapack.cpp add before line 36621:40
blackburnsame before 37221:41
blackburnplease ;)21:41
@sonney2kwe really need him for the ruby stuff :(21:41
blackburnand run example again21:41
@sonney2kpretty please21:41
blackburnI need to know where 20th parameter set to wrong value21:42
@sonney2kblackburn, wait info is a ptr21:42
blackburnoh yes21:42
@sonney2kdo you mean *info ?21:42
@sonney2kon your command21:43
blackburnhmm okay21:43
@sonney2kI really don't understand why the CPU fan is at 250 RPM even under full load21:44
@sonney2kon the other hand the cpu is at <60 C21:44
@sonney2kso probably it just doesn't care21:44
@sonney2k(my notebook is at 90 C ...)21:44
blackburnsonney2k: what is "int" size on your machine?21:45
@sonney2kIIRC int is always 32bit21:45
@sonney2klong int would be 6421:45
blackburnI can't understand why it can fail on my machine21:45
blackburnit seems andreas uses 64 too21:45
@sonney2kit cannot fail you mean?21:45
@sonney2kit does not fail?21:46
blackburnoh sh21:46
blackburnon my machine everything is ok21:46
blackburnsonney2k: could you please SG_SPRINT after "lwork = (int)work[0]"?21:47
@sonney2kblackburn, line?21:48
blackburnor 368 now21:48
@sonney2kblackburn, why is work a double* ?21:49
@sonney2kbut lwork an int?21:49
blackburnLWORK   (input) INTEGER21:49
blackburnsonney2k: LWORK is a length of WORK21:49
@sonney2klwork = (int)work[0];21:50
@sonney2kbut that converts the double to an int21:50
blackburnsonney2k: on the first call it determines the optimal work size21:50
blackburnand stores it in work(1)21:50
blackburnor work[0]21:50
@sonney2ksure about that?21:50
blackburnWORK    (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) *          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.21:50
@sonney2kok but then that is the error21:50
blackburn20th argument is wrong21:51
@sonney2kyou need to do lwork = *((int*) work);21:51
blackburntry that21:51
blackburnsonney2k: so why lwork = *((int*) work)? any difference?21:56
@sonney2kblackburn, yes it works now21:56
@sonney2k(my machine is not that fast...)21:56
blackburnso examples works now?21:56
blackburnsonney2k: I'll fix all the similar errors21:57
blackburnand commit21:57
@sonney2kblackburn, I hope you understand what the error was21:57
@sonney2kthen better try hard21:57
@sonney2kit is important21:57
@sonney2kyou have double* work;21:58
@sonney2kso what will work[0] give you ?21:58
blackburnfirst double element21:58
@sonney2kand then when you do (int) of it?21:58
@sonney2kint of the *double* element21:58
blackburnI got work[2]21:58
blackburnin means of integer array21:58
@sonney2kint of the first 64bit double21:59
blackburnyeah I see21:59
@sonney2kso the fix is clear now?21:59
blackburnyeah seems so22:00
@sonney2kok good22:00
@sonney2kI had a long chat session with baozeng today about some csharp marshalling and the errors he made were very similar22:00
blackburnoh it was hard I guess :D22:01
@sonney2kread the logs22:01
CIA-3shogun: Soeren Sonnenburg master * r0fcedaa / examples/undocumented/python_modular/graphical/ : remove all_evaluation -
@sonney2khemisphere data is gone again22:02
@sonney2kcould it be that you didn't commit data?22:02
@sonney2kblackburn, in the shogun repository?22:03
@sonney2kI mean like git commit data ?22:03
blackburnsonney2k: yeah may be22:04
@sonney2kplease do so22:04
@sonney2kthen I can git submodule update22:04
shogun-buildbotbuild #114 of csharp_modular is complete: Success [build successful]  Build details are at
@sonney2konly ruby and java left22:05
CIA-3shogun: Sergey Lisitsyn master * rf1cc27a / data : Data update -
blackburnsonney2k: should be ok now22:06
@sonney2kblackburn, yes indeed22:07
@sonney2kI don't get the example to work though22:07
blackburnwhich example?22:08
@sonney2kwhen commenting the 3d stuff22:08
@sonney2kI still have22:08
@sonney2k    swiss_roll_fig.scatter(X[0], X[1], X[2], s=10, c=tt,
@sonney2kTypeError: scatter() got multiple values for keyword argument 's'22:08
@sonney2kso I guess it needs a new matplotlib22:08
blackburnsonney2k: yes it is 3d scatter22:08
blackburncomment it out22:08
blackburnsonney2k: not really there is a way to do that for your matplotlib22:09
blackburnI don't remember how22:09
@sonney2kblackburn, would be cool if you could do a try: except: around this22:09
@sonney2ksuch that not all figures fail22:09
blackburnsonney2k: no, it is possible to do that even in old matplotlib22:09
@sonney2kshh ok22:09
@sonney2kahh ok22:09
blackburnbut I need to fix things in lapack22:10
@sonney2kso we now have to fix 2 graphical examples22:10
@sonney2k4 java bugs22:10
@sonney2kyour lapack issue (basically resolved)22:10
@sonney2kand the ruby examples22:10
@sonney2kthen everything works \o/22:10
blackburnsonney2k: I should add lapack test for libshogun22:13
blackburnlike I did for arpack22:13
@sonney2kI don't understand?22:13
@sonney2kyou mean some mathemetics_lapack.cpp ?22:14
blackburnsonney2k: now it fails on my machine?!22:19
@sonney2kblackburn, where?22:20
blackburnlapack issue22:20
@sonney2kyeah but not in the same line right?22:20
@sonney2kshow me the code please22:20
blackburnah no22:20
blackburni did22:20
blackburnlwork = *((int*)work);22:21
blackburnsonney2k: but works in the old way22:22
blackburnprobably better to use double work1 variable22:22
blackburnlike you did in dsyev22:22
@sonney2kbut my code is wrong too22:23
blackburnsonney2k: ?!22:26
@sonney2kthe (int) work1;22:26
@sonney2kit has the same problem22:26
blackburnnot possible22:28
@sonney2kblackburn, could you please print out the returned integers?22:28
@sonney2kin dsyevr / dsyev22:29
@sonney2konce with the (int*) thing and once with the int(work1)22:29
blackburnbetter dsygvx22:31
blackburnDSYGVX lwork=2150022:31
blackburn21500 for WORKAROUND ARRAY? :D22:32
@sonney2ksounds good22:32
@sonney2kwhat else?22:32
@sonney2ktemp buffer22:32
@sonney2knot workaround22:32
blackburnbut larger than I thought22:32
blackburnit is bigger than matrices..22:32
blackburnah no22:33
blackburni'm crazy22:33
blackburnsonney2k: if I use *((int*)work) I got 022:33
@sonney2kok so that is obviously not the intended value22:34
@sonney2kbtw I find22:34
@sonney2kthey also use (int) work1;22:34
blackburnI will try with work122:34
-!- f-x` [~user@] has quit [Remote host closed the connection]22:35
@sonney2kblackburn, but that is the same...22:35
blackburnno it is not22:35
CIA-3shogun: Sergey Lisitsyn master * r139c42a / src/shogun/mathematics/lapack.cpp : Attempt to fix DSYGVX -
blackburnsonney2k: could you please update and try?22:37
blackburnokay so it should be the common way22:39
@sonney2kblackburn, now I see what you did wrong: you called SG_FREE and then accessed the content of work[022:39
@sonney2k SG_FREE(work);22:40
@sonney2k367 22:40
@sonney2k-  lwork = (int)work[0];22:40
blackburnokay it is better anyway22:40
@sonney2kblackburn, ok now please continue with java - if you happen to get in touch with serialhex ask him for rub rub ruby :)22:43
@sonney2kmy ruby is not good enough - I have a couple of Q's22:43
blackburnwe should call him as russian kids call ??????????22:44
blackburni'm not sure you know wtf it is22:44
@sonney2k 22:44:52 up 1 day,  2:58,  1 user,  load average: 8.82, 10.51, 10.1422:45
@sonney2kthe buildbot is working22:45
@sonney2kbut cpu temp 54 - highest22:45
@sonney2kblackburn, no I don't know22:46
blackburnimpressive load22:46
@sonney2kblackburn, I find interesting images when I google for it though
blackburnyeah :D22:46
@sonney2kis it the male or female one?22:48
blackburn?????????? is some kind of wife for ??? ?????22:49
blackburn??? ????? is ~ santa claus22:49
@sonney2kyeah that I knew22:49
@sonney2kmoros == snow?22:49
blackburnsorry it is his daughter22:49
blackburnmoroz is hmm22:49
@sonney2kuhh ohh22:49
blackburnhard cold22:49
@sonney2kmy russian became totally useless22:50
blackburnthat's crazy but I wrong again22:50
@sonney2kyours too?22:50
blackburnsnegurochka is a GRANDdaughter22:50
blackburnsonney2k: ???? = snow22:51
blackburnsonney2k: you was near22:51
@sonney2kI guess one should sue you for the false claims :D22:51
@sonney2kanyway please go on22:51
blackburnwith lapack or snegurochka? :D22:52
blackburnsonney2k: could you please make a screenshot of output of swissroll?22:53
blackburnit probably could help to avoid some crazy shit after22:53
blackburnlandmark isomap is crazy but it is random22:55
blackburnsonney2k: need your help again22:55
blackburnor your machine, the same22:56
blackburnsonney2k: please run
blackburnoh no23:00
@sonney2kyes nice 3d figure23:08
blackburnwhy you haven't answered 10s before :D23:08
blackburninstalled old matplotlib23:09
@sonney2kdid you punish you bear23:09
shogun-buildbotbuild #136 of cmdline_static is complete: Failure [failed test_1]  Build details are at  blamelist: blackburn91@gmail.com23:10
blackburnsonney2k: the problem is I can't combine 3d and 2d plots in matplotlib 0.99..23:11
blackburnah wrong23:11
blackburnwhat the f I'm saying23:12
shogun-buildbotbuild #137 of cmdline_static is complete: Success [build successful]  Build details are at
blackburnuh it is hard to combine versions heh23:20
blackburnsonney2k: GOT IT23:27
blackburnshould work now..23:27
CIA-3shogun: Sergey Lisitsyn master * rcc0f888 / (2 files): Hack made hemisphere and swissroll work both in matplotlib 0.99 and >1.0 -
blackburnsonney2k: could you please try?23:34
@sonney2kit opens 2 figures23:45
blackburnsonney2k: unfortunately I can't combine 3d and 2d so for 0.99 - with two figures only23:46
blackburnon >1.0 it would work better23:46
blackburnstarting java examples fixing23:47
--- Log closed Sat Aug 27 00:00:29 2011